# tinker.types.LoraConfig

## *class* [**tinker.types.LoraConfig**](https://github.com/thinking-machines-lab/tinker/blob/main/src/tinker/types/lora_config.py#L8)(*[StrictBase](https://github.com/thinking-machines-lab/tinker/blob/main/src/tinker/_models.py#L76)*)

**Fields:**

- [**rank**](https://github.com/thinking-machines-lab/tinker/blob/main/src/tinker/types/lora_config.py#L9) (*int*) – LoRA rank (dimension of low-rank matrices)

- [**seed**](https://github.com/thinking-machines-lab/tinker/blob/main/src/tinker/types/lora_config.py#L12) (*Optional[int]*) – Seed used for initialization of LoRA weights.

  Useful if you need deterministic or reproducible initialization of weights.

- [**train_unembed**](https://github.com/thinking-machines-lab/tinker/blob/main/src/tinker/types/lora_config.py#L18) (*bool*) – Whether to add lora to the unembedding layer

- [**train_mlp**](https://github.com/thinking-machines-lab/tinker/blob/main/src/tinker/types/lora_config.py#L21) (*bool*) – Whether to add loras to the MLP layers (including MoE layers)

- [**train_attn**](https://github.com/thinking-machines-lab/tinker/blob/main/src/tinker/types/lora_config.py#L24) (*bool*) – Whether to add loras to the attention layers